Task 1: Setting up the sample
You build and deploy two machine learning models in Watson Studio. Then, you import a
sample decision service into Decision Designer and connect it to the deployed models.
- Follow the instructions available in AutoAI tutorial: Build a Binary Classification Model to create
two AutoAI experiments using the following datasets:
customer_churn_data.csv: In Configuration details, select No for the option to create a Time Series Forecast. Then, chooseCHURNas the column to predict.customer_LTV_data.csv: In Configuration details, select No for the option to create a Time Series Forecast. Then, chooseLTVas the column to predict.
- When the pipeline generation process completes, save one pipeline as a model from the resulting candidates.
- Deploy the saved model following the instructions available in Task 1: Creating an empty deployment space in Watson Studio.
- Import the Machine learning sample - Customer loyalty sample decision
service into Decision Designer:
- Sign in to Decision Intelligence Client Managed Software using your instance credentials.
- Create a decision automation.
- Click New decision +.
- Click Discovery tutorials and select Machine learning sample - Customer loyalty. Then, click Import.
- Open the Telco Retention decision service. It includes:
- Retention Offer: A decision model that recommends the best retention offer to prevent churn.
- Customer Lifetime Value: A predictive model that estimates customer lifetime value.
- Customer Churn: A predictive model that predicts customer churn risk.
- Follow the instructions available in Step 2: Defining a machine learning provider to create a machine learning provider.
- Follow the instructions available in Task 3: Uploading a PMML file and connecting a predictive model to:
- Connect the Customer Churn predictive model to the model deployed using
customer_churn_data.csv. - Connect the Customer Lifetime Value predictive model to the model
deployed using
customer LTV data.csv
- Connect the Customer Churn predictive model to the model deployed using